XPoint: Leaked Intel specs reveal 'launch-ready' SSD – report Leaked Intel Optane SSD specs show much higher endurance and lower latency than NAND SSDs and good, but not great, IO performance. M.2 format 16GB and 32GB caching Optane drives have already been announced, but no add-in-card/SSD-type drives yet. However, the Chinese TechBang website says it has seen a data sheet for an Intel …
Haven't worked on data integrity features for almost a decade, but 4228 would allow a 4K sector to have a T10 DIF field on each of 8 512 byte sub-sectors.
More likely it fell out of the implementation (which allows 512 byte sectors with T10 DIF) and got put on the data sheet as a potential differentiator.
Less likely: someone is playing with a capabilities architecture, access control tags, or the like, and simply wants to be able to reserve space at the sector level for future use. This is exactly what the 520 byte sectors offered, used in proprietary ways ~20 years ago in high end arrays like Symmetrix to ensure data integrity.
The article points out that some of the specs are not quite as good as some competing products.
I would counter that by pointing out that this is the first product of its type; it's unlikely they've hit the peak capabilities of XPoint first time round, so the fact that it's right up there with the top performers is pretty impressive, and bodes well for future generations of the product.
The best thing about 3D XPoint more than raw performance is endurance. 30 full drives writes per day means that a 3D XPoint drive is far more reliable than the best Enteprise SLC NAND SSDs and on par with or better than most Enterprise class Hard Disk Drives.
While MLC, TLC and upcoming QLC NAND SSDs are a huge mess, completely unreliable for any use other than caching on SAS RAID controllers.
3D XPoint SSDs will be able to replace Hard Disk Drives in a 5-10 years time and the unreliable NAND technologies will quickly become obsolete as they deserve.
write endurance hasn't been a factor for most workloads for years now already on enterprise systems.
Many(most?) vendors already offer unconditional warranties 3-5 years. Looking at the first AFA my company bought about 26 months ago (while it doesn't do a lot of traffic relative to what the system is capable of, it is running at about a 90% write workload- databases and close to 1000 VMs), the oldest SSDs are down to 97% of their expected lifespan for endurance. All of these SSDs if bought on the regular market would be sold as "read intensive". At this rate I suppose I may be lucky if the SSDs break below 90% of lifetime before the unconditional warranty expires(5 year).
The 2nd round of SSDs was added almost a year ago, and the third round in November I think, those two sets of SSDs are at 100% of life remaining.
NAND SSDs are not reliable.
Not an issue Enterprise wise ?
Only idiots and fools would spend millions on NAND SSD units expecting those to be reliable.
"they just wear out" ? Yeah sure.. just quickly re-buy all of those and keep spending 7-10 times the money that would be needed using Hard Disk Drives instead... that must be so smart indeed.. maybe for managers allowing these things in IT depts on purpose to steal money from investors offering a lower quality service overall.
NAND SSD's used in enterprise arrays are significantly more reliable than disks. There are good theoretical reasons for this and its supported by a large corpus of data collected across multiple years and thousands of installations and hundreds of thousands of devices.
Of course, this reliability may be a lot better than throwing a small number of drives into a server because the drives that do get used in enterprise arrays undergo extensive qualification, and the write workload in a well designed array will be evenly distributed across at least 12 and more often 48 devices means the endurance is further magnified.
Using a worst case examples for All Flash FAS from Netapp (who is my employer), you can write at least 90 Petabytes of data to a shelf of 24 SSDs before the drives begin to wear out
That equates to having a write workload of over 100,000 4K IOPS every second of every day for over eight years. The bigger the drives get, the better these figures get. To date, zero drives have needed replacement for drive wear, and the last time I looked, nobody was getting anywhere close to needing a replacement due to drive wear within the next five years.
If the drives do wear out during the service period (typically about 6 years before it makes sense to replace them to take advantage of better density and power) they get replaced at zero cost, regardless of the workload that was sent to them. If your vendor is forcing you to re-buy the drives because they are wearing out, then you need to look for a new vendor.
At this point the write endurance for the new Octane drives looks impressive and will be useful for things like small write intensive caches where you only have a couple of devices, and can't take advantage of wear levelling across a large number of large drives. The latency though .. 10 microseconds .. if you can make the I/O path to the drives fast enough to take advantage of that, then that has some really interesting implications, especially for RAM sizing and the next generation of "in-memory" database technology.
In all this discussion, don't forget this alleged product is using the new 3Dxpoint memory technology in a device which is form, fit, function replacement for one based on today's NAND flash.
This allows the new technology to be shaken out, and if an unfixable problem occurs, the warranty cost is limited to giving a bounded number of customers flash products to replace these new ones.
The best and highest use of byte addressable storage class memory is, of course, presenting a DAX space directly to the application, with no storage stack. This brings with it the host of issues discussed in the SNIA NVM programming model documents. The advanced topic here, of sharing byte addressed storage class memory between servers, is what the Gen-Z memory semantic fabric is about. So any characterization of 3Dxpoint as an insigificant performance boost, based on an early product that emulates today's flash drives, has missed the point.
@FStevenChalmers
It is just me or is calling it DAX an obscure deep space nine reference .. something that lives inside another host .. and who’s memory transcends and significantly exceeds the life of that host .. notably DAX a TRILL (Transparent interconnect of Lots of Links) and can take a variety of external forms ???
I know it’s probably just me.
But if someone comes up with a “Global Object Access and Universal Link Descriptor (GOA’ULD)” .. to access DAX then I know someone’s taking the piss
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
